Sunday, July 26, 2015

Facts have Challenge Ratings

No comments:
http://slatestarcodex.com/2015/07/23/the-general-factor-of-correctness/

Alexander neglects the challenge rating of facts. Every issue has a CR which your general epistemic competence g is checked against. If g >> CR, you get the correct answer. (Some caveats.) The opinion of a low g-rating mind on a high-CR question is determined by tribe and chance. Obvs if there's a tribal signal to display they'll display it, otherwise it's random.

Scales are arbitrary but I pegged journalists at 1-2 which puts Moldbug's g rating around 90. Don't confuse it with a percentage, I'm aiming for 120. I would also peg Scott Alexander's but he lies too much, I can't be arsed to work out which of his beliefs are sincere and which he's just pretending to not know.

1. Five Thirty Eight is down the night before an election, so you search for some other good sites that interpret the polls. You find two. Both seem to be by amateurs, but both are well-designed and professional-looking and talk intelligently about things like sampling bias and such. The first site says the Blue Party will win by 5%; the second site says the Green Party will win by 5%. You look up the authors of the two sites, and find that the guy who wrote the first is a Young Earth Creationist. Do you have any opinion on who is going to win the election?
In general it depends on the relative CR ratings, which I don't know off hand.  Even if Creationism in fact has a low CR rating, this is not evidence the Green Party site knows what it's talking about; what Alexander thinks is 'talking intelligently about sampling bias' can and is usually learned by rote, and thus useless. They will have to show their work, in which case I can evaluate their work directly.

Creationism is red tribe, not blue, so it may be a declaration against interest, which gives a bonus to g estimate.
2. On the bus one day, you sit next to a strange man who mumbles about how Bigfoot caused 9-11 and the Ark of the Covenant is buried underneath EPCOT Center. You dismiss him and never see him again. A year later, you see on TV that new evidence confirms Bigfoot caused 9-11. Should you head to Florida and start digging?
Depends on the cost. It doesn't lower the probability but it doesn't raise it much either. If you were going to Florida with backhoes anyway...well, why not? If you're a regular person, no. People who mumble in public are showing ignorance of social rules, which has a very low CR. (Do I want fractional CR or do I want it logarithmic?) It is most likely caused by the fact that strange things remind you of other strange things, and thus when you think of solutions to strange thing A), strange thing B) is in the availability bias slot.

There's also the very likely case that the TV is mistaken; as above, journalists are lucky if they breach g level 2.

There is a very small probability that they're ignoring social rules because they have atypical values, and thus aren't being irrational. Never personally observed a real specimen, but they should theoretically exist.
3. Schmoeism and Anti-Schmoeism are two complicated and mutually exclusive economic theories that you don’t understand at all, but you know the economics profession is split about 50-50 between them. In 2005, a survey finds that 66% of Schmoeist economists and 33% of anti-Schmoeist economists believe in pre-Clovis settlement of the New World (p = 0.01). In 2015, new archaeological finds convincingly establish that such settlement existed. How strongly (if at all) do you now favor one theory over the other?
Given the past performance of economists-who-make-mainstream-news-so-that-we-know-there's-a-controversy, most likely they're both wrong anyway.

But neglecting that (obviously higher-than-SA's-g-rating-CR) fact, without understanding Schmoeism to a reasonable degree you won't know its CR. If Schmoeism's CR is lower than Clovis, then it is evidence for Schmoeism roughly to the same ratio the CR is lower. If CR is 0 then Schmoeism is all but guaranteed; if it's 1-2 points lower, then there's no evidence either way; and if it's 20 or more points the wrong side is 75% likely all shills.

If Shmoeism's CR is significantly higher than Clovis they don't even understand the question, and it might be unreasonable to assume they're arguing about it.
4. As with 3, but instead of merely being the pre-Clovis settlement of America, the survey asked about ten controversial questions in archaeology, anthropology, and historical scholarship, and the Schmoeists did significantly better than the anti-Schmoeists on 9 of them.
With more questions, the odds that one of them has high CR is better.
Of course real scholars go and look when they can, rather than trying to hack weird probabilities. If Clovis' CR is reasonably high - which is the only way a real scholar wouldn't already know the answer - then the odds that any of them have high-enough CR is still negligible.

--

The fact that facts have a challenge rating is probably a high CR fact. Estimate 105 +/- 25.

The only person I've ever noticed to buck the CR trend is Elizer Yudkowsky, who bucks it in both directions, so I'm not sure what to make of it.

The CR of religion is well above 120. Most likely 150-200 region, but could easily be 300 or higher.

Thursday, March 5, 2015

Piketty Asserts the Logical Impossible

6 comments:
There are many ways in which Piketty is wrong. However, it seems nobody has noticed it is logically impossible for his primary empirical assertion to be correct. (There is no true a priori statements, but given that, I think the term can be re-purposed for things like this.)

As per the above link, GDP growth is asserted to be, rounded, 1% and return on capital is 5%. If this were true, non-capital forms of income would see decline, and hit zero quite quickly due to exponential compounding.

Say 50% of GDP is return on capital and 50% is other forms of income. Say GDP grows 1% and capital return is 5%, so capital amounts to 1000% of GDP. Let's just call it $1000 because why not. So GDP year 0 is $100.

Next year capital is $1050. GDP is $101. Capital return is still 5%, so $52.50. Call it 53.
Income is now $48. Capital is now $1103

GDP $102.01. Again I'll drop the cents from income and capital - $47 and $1158

GDP is $103.03 Income - $45 Capital - $1216

$104.06 || $43 || $1277

$105.10 || $41 || $1341

Etc. After only five years, income is now down nearly 20%. If population has grown, most would be dramatically poorer. In fact, everyone has been getting richer. Eventually income will hit zero, and capital returns will begin tanking. Though of course income = zero means everyone starves and the economy collapses entirely long before that.

Obviously this is slower if return on capital is a lower portion of total GDP, but it is logically impossible that return on ALL capital could be more than GDP growth for hundreds of years. If he hasn't fudged his numbers, all he's proven is that GDP is a worthless measure.

(Indeed it's slightly worse than this - if any of that income isn't spent, it becomes capital and reduces future income even further.)

So at 50%, non-capital income sees 3.3% decline.
Let's say capital is only 10%. $200.
$101 || $90.5 || $220.5

Income sees 0.6% growth in the first year. Capital is still, by definition, seeing 5%. Within decades non-capital income will stagnate. By year 10 it's already down to 0.3%. Imagine your final, year-65 retirement wage being less than 15% higher than your first job.

This non-capital decay only doesn't happen if capital is an utterly negligible part of the economy.

Wednesday, March 4, 2015

Unproductive Moral Fisking

1 comment:
This will learn me to stop clicking SSC links.

The consistently one attempts to adhere to an ideology, the more one's sanity becomes a series of unprincipled exceptions.

Neon Hillism: keep the exception and sacrifice the principle. It's not necessary to understand what you're doing all the time. If you can work out a new principle that doesn't fail, great. Otherwise, Accept your Ignorance. Hasn't failed on me yet.

In reality it’s pretty hard to come up with way of valuing animals that makes this work.

If by 'hard' you mean 'super easy.'
In fact animals have zero moral value. I still want them to not suffer, though. This value isn't in any way different from my value of not dying or finding ice cream tasty. The market will (eventually) provide an efficient balance of animal-not-suffering to ice cream once it is allowed to do so. That's what it's for. Just don't forget that animal suffering only has valence through the dire apes who care about it, which means it matters exactly to the extent the dire apes care about it.

But going from “just my community” to “also foreigners” is a difficult step that’s kind of at the heart of the effective altruism movement.

I have worked out what's wrong with this. It's basically statistical unsophistication. Not all cars are worthy transport, and it correlates weakly with firm and country of origin. Simlarly, not all people are worthy of concern, and it correlates weakly with ancestry and country of origin. Turns out this test is also super easy: will they show you reciprocal concern? If so, then be nice. If not, fuck 'em.

allowing starving Third World people into the circle of concern totally pushes out most First World charities like art museums and school music programs and holiday food drives. This is a scary discovery and most people shy away from it.

Which is fine, because it's wrong.
Letting die is not murder. My paragraph above is about who it's wrong to murder, e.g, not animals.
Who it's wrong to let die: anyone you've promised to not let die. Starving third-worlders can no more legitimately demand resources for free than anyone else. They claim they're starving? Fine, I claim to be a utility monster. "That's not reasonable!" Oh, we're using common sense now? (Special pleading.)
Further, as frequently demonstrated, first-worlders are quite willing to trade for resources on very favourable terms for the third-worlder. This would simply not be an issue if state coercion were not involved.

It is Gnon's judgement that third worlders will starve until they stop driving full tilt into starvation situations. Coercive first world aid cannot revoke Gnon's judgement, which is half the teleological reason it only ends up destroying local textile manufacturing, enriching revolting tyrants, and causing the AIDS plague.

No kind of free aid can revoke Gnon's judgement. Trade, however, can, though the intricacies of how this works are beyond the scope here. Though it's also relevant that if you were starving, no third-worlder would help you, even if they easily could.

I’d lend it to him, but I wouldn’t give him exactly half my money no-strings-attached on the grounds that he is exactly as important to me as I am.

Because you can't trust your brother as much as yourself. If he would likewise give you half his money, then there would be no need to keep separate bank account at all, as he would weigh his values exactly as heavily as yours. He won't, though, out of ignorance at least as much as out of selfishness and hypocrisy. This kind of principle only gives advantage to anyone claiming to be a utility monster.
Which I guess is fine. The utility monsters eat all the insane moral-signallers, then I hunt the utility monster, and sanity returns.

We can argue about what 'should' happen in this case as long as we like, but it won't change what will happen, which is Gnon will harshly judge anyone who treats the selfish as unselfish. The action symmetry is empirically not there.

my friends. We all raised a lot of money to help Multi

In case anyone thinks SSC is truly a symmetric, disinterested moralist, Multi has explicitly called for my death in particular. They are not banned from SSC, whereas I'm maybe one indiscretion away from being banned. Benefit of the doubt: not deserved.

I suppose, though, this is SSC insulting Multi and complimenting people like me. They are not a threat, no matter how vile their rhetoric. I am a threat, no matter how polite mine is. (It still encourages the vile.)

well, I know myself better than I know my brother, and I know Multi better than I know strangers, so I’m more effective

Compare to two quotes up, and remember the cowpox of doubt.

I don’t think I can make a principled defense of doing this. But I think I can claim I’m being unprincipled in a meta-consistent and effectively sanity-protecting way.

Turns out insane beliefs are sanity-destroying if you try to take them seriously. Who knew.
I dunno. If I found out one of my beliefs was paying 'rent' in the form of attacking my sanity, I would question its rationality. I mean, this is kind of an obvious reaction, right?

Monday, February 9, 2015

Weapon Cryto-lock Design

1 comment:
Needs either an unlock signal, or a dead man's switch that holds a lock shut. In either case, subject to jamming.

If a unlock signal is jammed, foreign powers can disarm your forces. Therefore it's a dead man's switch. This means rebels can put up sovereignty umbrellas. Therefore, locks only make rebellion more expensive, hardly impossible.

Still, front door locks don't make breaking in impossible, merely harder, and they're cost-effective. Perhaps the crypto-lock can use tricks to make it worse, like sharing a channel with radio chatter, meaning jamming the switch also jams your own comms.

There's an issue with small arms. They aren't hard to make yourself. Still, an M16 is not basement forge tech, and it puts up another couple barriers to rebellion. Also miniaturizing the locks to that extent greatly increases the malfunction risk, and the last thing you need is less reliable M16s. You can still lock the armoury door, though. Having a chance to shoot the rebels while they're jimmying the lock can make all the difference.

Futher, the actual trouble with locking small arms is a side issue when modern destruction is mainly wrought by planes and howitzers. If you can't win against a rebel army lacking artillery, then you richly deserve to lose. Even if it gives their heavy guns reliability issues, it will be cost-effective. The loyalists can destroy the gun, or destroy the jammer. This additional option would help tremendously in a real firefight.

There's also the issue that jammers can't, by definition, be hidden. If they're working they can be triangulated and targetted. This is perhaps a reason to avoid fibre optic lines, even though their jam-analogue is harder to effect.

Finally, a ruler can take the spirit of crypto-locks and run with it. Adding several layers of technological replacements for warm fuzzies would have multiplicative effectiveness.

Wednesday, February 4, 2015

Fertility is a Purpose Problem

No comments:
Turns out I'm interested enough in this to do it properly. Hopefully the following isn't too narcissistic, but I would do it regardless.

For my anti-confirmation-bias impulse, I will use the fact that "The Cathedral did it1" is getting boring as an explanation. Surely, just by statistical chance, it wouldn't always be their fault... In any case, today's pathology is the Cathedral's practice of preventing hobbits from satisfying their urge to join and further a higher purpose.



So, independent corroboration. Hopefully none of these read my comment. I certainly allowed it to get long enough to repel them, so it's likely. To be fair I should also look for opposing corroborations, to get a good baseline rate, but today I don't feel like doing the other side's work for them.



"Singapore shares a planet with America, the poor bastards. As a synecdoche, I expect Singaporean women still think having a career makes them more attractive to men."
vs.
"consider the case of East Asia: the relevant change was from property being held in common by the family and economic decisions being made by the family to an individualist basis for property and economic decision making."
and
"The market hasn’t spoken, gender equality is a completely artificial State-enforced policy."


In particular, the state changed the security arrangements, but did not change prices or security procedures. They changed to treating individuals as if they were male-headed families, which, shockingly, turns out to be irrational. Luckily they can extract rents coercively, or they would have gone out of business; it's pretty bad when your security monopoly goes out of business. Chaos = Δpower, after all.


"nobody but the Islamists alieves in God anymore. Lacking a society-wide purpose"
vs.
"It’s harder than I thought to get fertility stats for Mormons, but it appears that they breed well above replacement level, their desired family size is 4 children, and that high income/education does not suppress fertility."

I'm still dubious about Mormons' alief in God, but they certainly alieve in a higher purpose. Curiously, I'm sure of that even though I can't put their purpose into words.

"It becomes impossible to reward child-having with status "
vs.
"Women try hard to do what society tells them will earn admiration. The prescription for the short term is to restore prestige and style to the role of wife and mother, not threaten to chain us to the stove, barefoot and pregnant, because we are too dumb to do anything else. When men of the right agree with gays, and view us as just breeders, something is very wrong."

No disrespect to Jim - it is necessary that someone forwards and defends the idea - but he is incorrect. The impulse to assert such things is usually a signalling effort, trying to show maturity through the capacity to affirm dark truths. Unfortunately, this isn't a truth, so it indicates instead immaturity. That said I don't know what's up with Jim in particular.
Not to let Alice off the hook - it is likely she affirms this because she's disgusted with being chained to the stove, rather than because it's true. It is almost certain women are too 'dumb' - that is, unmale - to do something, and not just warfighting, and Teller wouldn't accept it. No disrespect here either - expecting non-aspergoids to be inhuman is foolishness - merely pointing out it's necessary to correct for these deficiencies.

"turns out there’s a fallback position within the human instincts, which is approximately consumerist hedonism."
vs.
"I personally suspect it has a lot to do with the hedonic treadmill. Even in Muslim countries, hardly hotbeds of feminism, wombs are going barren. I also recall some study showing a strong correlation between access to television and fertility." 

I'd say SGW needs to train intuition reading better, but regardless this is what intuition convergence looks like. System 1 is bad at communicating details so this is where to bring in system 2.



All this said, I still think there's a well-deserved bucket of cold water around here somewhere. I also think it's Land's idea and I'm going to delegate to him the task of finding it properly. 



[1] In the bedroom with the pipe.

Thursday, December 11, 2014

Confirmed Predictions II

No comments:
I attempted to increase my epistemic competence. My predictions started out generously at 60% right. Now I can casually assume they're right, so it seems to have worked.



Of course this is privileged information. I could easily have found the data and then pretended to have made the prediction. Indeed the match is uncannily close, given I spent less than a second before coming to a conclusion - and uncanny means literally/prosaically incredible. The idea is to point at what to look for in yourself, rather than for you to believe I did in fact do this. Secondarily, when I say one of my predictions came true, I mean I observe something like this.

It's oft repeated that nicotine has a lethal dose of 50mg. When I first saw this, today, I assumed they had misread the units - surely, that's supposed to be mg/kg? Google gave me the CDC, which I assumed had dropped the /kg, as it uses mg/kg everywhere else, including the immediately following example parenthetical. On being prodded, I looked further down the search results and found this and this, showing the myth is definitely out there in the wild. Because the CDC was inconsistent about units, I predicted that someone had dropped the units, and subsequently been seminal. (And that given that the human dose is rated as exactly the rat dose, it's probably basically a guess, not even informed by clinical case studies.)

Then I followed the link in that latter link.
"Some of these effects resemble typical symptoms of nicotine overdosing, but 1–4 mg of oral nicotine will certainly not evoke the severe adverse effects described, such as clonic seizures and loss of consciousness."
So either the dose was much higher than listed - 3mg/kg, not 3mg - or their chemical supplier screwed up and it wasn't nicotine. (Check date; credit goes to serendipity.) However, Mayer is not saying that those symptoms are unlike nicotine, instead explicitly saying they are like nicotine. (Wikipedia confirms.) It very strongly suggests the dosage unit was typoed.

The person who first dropped the units was Rudolf Kobert, who published in 1906, "in accordance with the severe symptoms evoked in several experimenters by 0.002–0.004 g it is certainly not going to be higher than 0.06 g." Explicitly guesswork.

That said, he was "a renowned pharmacologist" and thus no more likely to make bad guesses in his field than I am. Mayer's cautious lower limit for lethal dose is 1 gram, whereas Kobert said, if we assume the typo theory, that it certainly won't be above 2.5 grams for an average male of 1900. It's all consistent. Getting a mistake consistent doesn't happen by chance.



On the other hand Mayer needs to hang around engineers for a while. Check for small problems before assuming there's a big one. It's cheaper. While he was still helpful, having to steelman everyone gets tiresome after a while. Either learn logic properly or stick to reporting data and leave the interpretation up to an expert. I learned it and there's no reason Mayer couldn't too. In the worst case, the point for Mayer was to shame self-experimentation, not to show anything about nicotine.



Note about unmistakeable evidence - technically we must consider that Kobert didn't in fact observe seizures. All we know is that Kobert reported seizures. Or do we? We must, technically, doubt that we've seen the report. Names get typoed too, etc. Pragmatically, the chance is indistinguishable from zero. Pharmacologists don't think they've seen a seizure when they haven't, and if he'd tried to lie he would have been caught. Finally, that we know we think we've seen the report is not pragmatically like 100%, not even lim approaches 100%, it's plain 100% likely to be true.

Tuesday, December 2, 2014

Tet Offensive and Fourth Generation Warfare

1 comment:
Even though the Vietcong were materially defeated in detail, they won the moral victory, says Robert Greene:

Within a few weeks, in all parts of South Vietnam, the Americans regained the upper hand, retaking control of Saigon and securing their air bases. The sieges at Hue and Khe Sanh took longer, but massive artillery and air bombardments eventually doomed the insurgents, as well as leveling entire sections of Hue.
After what later became know as the Tet Offensive was over, Westmoreland likened it to the Battle of the Bulge, near the end of World war II.  There the Germans had managed to surprise the Allies by staging a bold incursion into eastern France. In the first few days, they had advanced rapidly, creating panic, but once the Allies recovered, they had managed to push the Germans back -- and eventually it became apparent that the battle was the German military's death knell, their last shot. [...] The entire Vietcong infrastructure had been wiped out.
[...]
But another viewpoint began to trickle in from home: the drama at the U.S. embassy, the siege of Hue, and the attacks on air bases had kept millions of Americans glued to their television sets. Until then the Vietcong had operated mostly in the countryside, barely visible to the American public. Now, for the first time, they were apparent in major cities, wreaking havoc and destruction. Americans had been told the war was winding down and winnable; these images said otherwise. Suddenly the war's purpose seemed less clear. How could South Vietnam remain stable in the face of this ubiquitous enemy? How could the Americans ever claim a clear victory? There was really no end in sight. 
American opinions polls tracked a sharp turn against the war. Anti-war demonstrations broke out all over the country. President Lyndon Johnson's military advisers, who had been telling him that South Vietnam was coming under control, now confessed that they were no longer so optimistic. In the New Hampshire Democratic primary that March, Johnson was stunned by his defeated by Senator Eugene McCarthy, who had galvanized the growing antiwar sentiment. Shortly thereafter Johnson announced that he would not run for reelection in the upcoming presidential race and that he would slowly disengaged American forces from Vietnam
The Tet Offensive was indeed the turning point in the Vietnam War, but not in the direction that Westmoreland and his staff had foreseen.